TIESFeaturedtiesmultilingualcodingmistralqweninstruct

OpenHermes Fusion — Mistral × Qwen TIES Multi-Model

A TIES merge of OpenHermes-2.5-Mistral-7B and Qwen2.5-7B-Instruct, targeting improved multilingual capability and coding while preserving Hermes instruction tuning. Sign conflicts resolved at density 0.7.

Author

teknium

Published

November 10, 2025

Last updated

January 20, 2026

Versions

2

Best Score

↑ 71.3

Stars

187

Mistral-7B-v0.1 · 7BQwen2.5-7B-Instruct · 7B
Share

Merge Lineage

Merge Lineage

2 source models

Source Models

Intermediate

Output

openhermes-qwen-fusion-7b

7B

TIES

Click any model node to open its Hugging Face page

Config YAML

openhermes-mistral-qwen-ties-v1.2.yaml
merge_method: ties
base_model: mistralai/Mistral-7B-v0.1
models: - model: teknium/OpenHermes-2.5-Mistral-7B
    parameters: weight: 0.6
      density: 0.7
  - model: Qwen/Qwen2.5-7B-Instruct
    parameters: weight: 0.4
      density: 0.75
parameters: normalize: true
dtype: bfloat16

Benchmark Scores

BenchmarkMergedOpenHermes-2.5Qwen2.5-7BΔ Best
MMLUtop
71.369.869.1+1.5
HumanEval
65.461.263.4+2.0
MT-Bench
7.87.67.4+0.2

Model Weights & Density — TIES

OpenHermes-2.5-Mistral-7B7B
weight
0.60
density
0.70
Qwen2.5-7B-Instruct7B
weight
0.40
density
0.75

How I Built This

Hugging FacePaper

Embed Badge

Add this to your Hugging Face model card to link back to this recipe.

MergeKitRecipe
markdown
[![MergeKit Recipe](https://img.shields.io/badge/MergeKit-Recipe-10b981?style=flat-square)](https://www.mergekit.com/recipes/openhermes-mistral-qwen-ties)

Version History

  1. v1.2latest
    ↑ 71.3

    January 20, 2026

    Increased Qwen density to 0.75 — better multilingual without degrading English

  2. v1.1
    ↑ 70.1

    November 10, 2025

    Reduced Qwen weight to 0.4 to prevent instruction format drift

Use this Model

Run, deploy, or interact with OpenHermes Fusion — Mistral × Qwen TIES Multi-Model directly.

MergeKit Cloud

Run this model on serverless GPU infrastructure — zero setup, pay-per-second.

RunPod Serverless

Select GPU

A10G · 24 GB~45s cold
A100 · 80 GB~30s cold
H100 · 80 GB~20s cold
Est. ~$0.001 / inference · billed per second
Coming Soon

Serverless GPU · Powered by RunPod

Reproduce Locally

Run this exact merge on your own machine in three steps:

bash
pip install mergekit
yaml
merge_method: ties
base_model: mistralai/Mistral-7B-v0.1
models:
  - model: teknium/OpenHermes-2.5-Mistral-7B
    parameters:
      weight: 0.6
      density: 0.7
  - model: Qwen/Qwen2.5-7B-Instruct
    parameters:
      weight: 0.4
      density: 0.75
parameters:
  normalize: true
dtype: bfloat16
bash
mergekit-yaml openhermes-mistral-qwen-ties.yaml ./output

Want to build your own merge?

Use the MergeKit config generator to build a YAML recipe visually — no code required.

Open Config GeneratorBrowse More Recipes